A React-based visual editor for designing and prototyping Pytorch neural network architectures. This tool allows users to drag-and-drop layers, configure parameters via a dynamic UI, and instantly generate the corresponding Python code.
- Visual Graph Editor: Build on top of React Flow, enabling intuitive drag-and-drop construction of architectures.
- Instant Code Generation: Compiles the visual graph into a valid, copy-pasteable PyTorch
nn.Moduleclass, including__init__andforwardmethods. - Shape Inference: Calculates and displays tensor shapes in real-time as you configure layers
npm install
# or
yarn installbackend dependencies:
python3 -m venv venv
source venv/bin/activate
pip install -r backend/requirements.txt- Start the frontend development server
npm run dev
- Start the backend server (for torchlens)
uvicorn backend.torchlens_server:app --host 0.0.0.0 --port 8000This project is a combined effort from DSG club and SDSLabs at IITRoorkee.